Inferring Ideological Ambiguity from Survey Data

Abstract

The paper presents a Bayesian model for estimating ideological ambiguity of political parties from survey data. In the model, policy positions are defined as probability distributions over a policy space and survey-based party placements are treated as random draws from those distributions. A cross-classified random-effects model is employed to estimate ideological ambiguity, defined as the dispersion of the latent probability distribution. Furthermore, non-response patterns are incorporated as an additional source of information on ideological ambiguity. A Markov chain Monte Carlo algorithm is provided for parameter estimation. The usefulness of the model is demonstrated using cross-national expert survey data on party platforms.

Keywords: Bayesian Item Response, Ideological Ambiguity, MCMC